Surface ozone

While individual CTMs are routinely compared with observations for specific scientific studies, comprehensive evaluations against observations in some regions of interest for HTAP remain limited. In the HTAP intercomparison, the model ensemble mean captures observed monthly mean surface O3

throughout the year over Europe and the western U.S. but overestimates it substantially during summer and early fall over the eastern U.S. where sites are influenced by major pollutant sources and over Japan where simulations are sensitive to the timing and extent of the summer monsoon [Figure 4.3, Fiore et al., 2009]. These results are consistent with observational comparisons during the

ACCENT/PHOTOCOMP study, which showed that, despite large inter-model differences, the model ensemble annual mean generally fell within 5-10 ppbv of observed values, except for South Asia where the models exhibited a systematic positive bias of 15-20 ppbv compared to the limited measurements available [Dentener et al., 2006; Ellingsen et al., 2008]. The seasonality in surface O3 was reproduced relatively well [Ellingsen et al., 2008].

140

Figure 4.3. Comparison of monthly mean surface O3 from models contributing to the HTAP intercomparison (grey lines; multi-model ensemble mean in black) compared with observed O3

(red; vertical lines denote ±1σ variability in the monthly mean over the different stations) from measurement networks over the northeastern U.S. (CASTNet), central Europe (EMEP) and Japan (EANET). Models are sampled at the nearest grid points to each station. [Adapted from Figure 2 of Fiore, A. M., et al. (2009), Multimodel estimates of intercontinental source-receptor relationships for ozone pollution, Journal of Geophysical Research, 114(D4): D04301.]

Reidmiller et al. [2009] conducted an extensive evaluation of the HTAP CTMs with the policy-relevant maximum daily 8-h average ozone (MDA8 O3) metric from the Clean Air Status and Trends Network (CASTNet) observations in the U.S. They found that the multi-model mean represents the observations well (mean r²=0.57, ensemble bias=+4.1 ppbv for all U.S. regions and all seasons), though individual model results vary widely. The models generally reproduce the observed day-to-day

variability well [e.g., Figures 5 and A5 in Reidmiller et al., 2009], with the strongest correlations between simulated and observed MDA8 O3 in the northeastern U.S. during spring and fall (r²=0.68) and weakest in the midwestern U.S. in summer (r²=0.46), but biases of +10-20 ppbv occur in summer over all eastern U.S. regions (as in Figure 4.3 for the monthly mean values). The seasonal and daily

variability is fairly well reproduced, but there is a systematic negative bias at high altitude sites in the mountainous western U.S. which are more influenced by free tropospheric air that contains larger contributions from foreign emissions (i.e., those outside the region) see Figure 4.4. More information for particular regions of the U.S. and for different seasons is provided in Reidmiller et al. [2009].

Figure 4.4. Comparison of maximum daily 8-hour averaged surface O3 (MDA8 O3) over different parts of the U.S. in spring 2001 from models contributing to the HTAP

intercomparison with observations from the CASTNet network. The model mean is shown in black (grey shading indicates ±1σ) and observations are shown in red. [Adapted from Figure 5 in Reidmiller, D. R., et al. (2009), The influence of foreign vs. North American emissions on surface ozone in the US, Atmospheric Chemistry and Physics, 9: 5027-5042.]

FINDING (ozone): Current global models reproduce the observed regional and seasonal

variability in surface ozone at most locations, demonstrating our ability to represent the key large-scale processes controlling the formation, transport and removal of ozone and its precursors.

However, significant discrepancies exist on shorter spatial and temporal scales indicating weaknesses in our representation of local- and urban-scale processes in current models.

141 4.2.3. Trends in surface ozone

Over recent decades, significant trends in tropospheric O3 have been detected from many observational platforms (aircraft, balloons, high-altitude stations, and satellites) over different regions and time periods. Taken together, these observations paint a complex picture with differences in the magnitude and sign of trends across parts of the northern hemisphere [NRC, 2010, and references therein; Royal Society, 2008b; TFHTAP, 2007]. For surface O3, measured trends are more consistent qualitatively, with general increases ranging from 0.1‒0.5 ppbv yr^-1 in the western U.S. and Western Europe, though some observational records suggest little change or a recent stabilization [see summaries of literature by NRC, 2010, Tropospheric ozone trends workshop, 2009; Royal Society, 2008b;

TFHTAP, 2007; Vingarzan, 2004]. Differences in observational sampling frequency and time periods combined with large interannual fluctuations in meteorology and emissions complicates both

interpretation of trends from measurements and reproduction of the magnitude of long-term O3 changes using models.

Because the HTAP model experiments were focused on a single year, no explicit evaluation of trends was included in the HTAP intercomparison. However, past studies with models that participated in the intercomparison have been evaluated against observed trends, with mixed results.

Under a major European modelling initiative, RETRO, 40-year simulations (from 1960 to 2000) were conducted with three global CTMs and evaluated with long-term O3 measurements [Schultz et al., 2007]. All models indicate an increase in O3 of 5‒15 ppbv over Europe, North America and East Asia south of 55°N over this period and capture much of the observed interannual variability [Schultz et al., 2007]. However, increases observed at central European mountain stations are not reproduced,

suggesting weaknesses in our understanding of source changes, including emissions and the influx of O3 from the stratosphere. More recent studies with two chemistry-climate models using new

assessments of historical emission changes show a positive trend in surface O3 at many locations, but this is generally somewhat lower than has been observed [Lamarque et al., 2010], as shown in Figure 4.5. Regional modelling studies over East Asia have shown a similar underestimate of observed surface O3 trends [Tanimoto et al., 2009]. Future effort is needed to reconcile modelled and observed O3 trends based on improved understanding of processes and emission changes before these trends can be reliably attributed to particular sources, or before we can definitively determine whether this inability to capture the observed trend reflects fundamental problems with current models.

Figure 4.5. Comparison of surface O3 trends (12-month running mean) from CAM-Chem (black line) and GISS-PUCCINI (black squares) against observations (red) at Zugspitze and

Hohenpeissenberg, Germany (left) and US Pacific Coast sites (right). [Adapted from Figure 5 in Lamarque, J. F., et al. (2010), Historical (1850-2000) gridded anthropogenic and biomass burning emissions of reactive gases and aerosols: Methodology and application, Atmospheric Chemistry and Physics, 10: 7017-7039.]

142

FINDING (trends ozone): Recent model studies have shown increases in surface ozone between 1960 and 2000, but the magnitude and regional variation in this trend does not compare well with observations. Observed increases at remote locations are generally underestimated in model simulations, indicating that precursor emissions and/or atmospheric processes are not represented well in current models.

RECOMMENDATION (ozone trends): Research is needed to characterise and explain the observed trends in surface ozone. Multi-year model simulations that account for changing emissions and meteorology are necessary to (1) explain and attribute changes, (2) put observed regional trends in a global context, and (3) critically test model ability to reproduce long-term composition changes. Proposals under the IGAC/SPARC AC&C activity to „hindcast‟ observed composition changes over past decades should be encouraged and supported.

4.2.4. Source attribution

Model studies of pre-industrial conditions demonstrate that anthropogenic emissions have had a large impact on tropospheric O3, contributing about 100 Tg (9 DU) to the global burden [Gauss et al., 2006; Prather et al., 2001], about 30% of the current burden of 335±10 Tg [Wild, 2007]. Around 40% of this change is due to increases in CH4 and the rest to changes in emissions of NOx, CO, and NMVOCs [Shindell et al., 2005]. The contribution of anthropogenic emissions to surface O3 is thought to be significantly larger, with O3 concentrations roughly doubling between 1850 and the present day [e.g., Berntsen et al., 2000; Lelieveld and Dentener, 2000]. Recent studies with a number of models suggest that anthropogenic emissions by year 2000 have caused global surface O3 to rise from a preindustrial level of about 17±3 ppbv to about 28±5 ppbv [Royal Society, 2008b]. However, there is large uncertainty in our ability to simulate the preindustrial atmosphere as the few surface O3

measurements available from the 19^th century reveal substantially lower concentrations than current models can reproduce [Mickley et al., 2001]. This suggests that models may be underestimating the anthropogenic contribution to surface O3. In light of this, we have high confidence that current surface O3 is significantly higher than preindustrial levels [NRC, 2010].

The HTAP intercomparison suggests that annual average surface O3 levels over the four HTAP regions are currently about 37±4 ppbv. Of this, 20-25% originates from the stratosphere, and a similar proportion is formed from natural precursor sources: lightning, soils, vegetation and fire, along with oxidation of natural CH4 [Lelieveld and Dentener, 2000; Sudo and Akimoto, 2007]. The

anthropogenic contribution thus typically exceeds 50% over these regions. About half of this contribution originates from sources over the region itself (8-10 ppbv) and about half is transported from sources outside the region. However, this simple attribution masks strong regional, seasonal and daily variability in both O3 abundance and in the contribution of different sources. We also note that while the contributions shown in Figure 4.6 are indicative of average northern mid-latitude conditions, they do not represent „background‟ conditions at any particular location.

FINDING (attribution of ozone): We have high confidence that emissions from human

activities contribute substantially to baseline ozone. Annual average surface O3 levels over the four HTAP regions are estimated to be 37±4 ppbv, of which over 50% may be attributed to anthropogenic sources. About half of this anthropogenic component originates from sources over the region itself (8-10 ppbv) and the other half is transported from sources outside the region. Another 20-25% of surface O3 originates from the stratosphere, and a similar

proportion is formed from natural precursor sources: lightning, soils, vegetation and fire, and from oxidation of natural hydrocarbons.

143

Figure 4.6. Source attribution for tropospheric O3 over the globe (left) and for annual mean surface O3 over the four HTAP regions (right) estimated by the report authors from source contributions in earlier published studies [Berntsen et al., 2000; Gauss et al., 2006; Lelieveld and Dentener, 2000; Sudo and Akimoto, 2007].

FINDING (attribution of ozone): We have confidence that human activities contribute to the observed positive trend in surface ozone, and that the anthropogenic component to hemispheric mean ozone levels has grown significantly since ~1850. However, there remain large

uncertainties in our estimates of the source attribution for these changes.

RECOMMENDATION (attribution of ozone): Dedicated model studies focussing on source attribution of ozone are needed to reduce the uncertainties in current estimates, and to better characterize the strong regional, seasonal and daily variability in both ozone abundance and in the contribution of different sources. Improved, observation-based constraints are needed to provide a critical test for these model estimates.

4.2.5. Source-receptor relationships for surface ozone

A number of published studies have estimated the influence of foreign sources on surface O3

in different parts of the northern hemisphere [TFHTAP, 2007]. These studies have applied a wide variety of techniques (described in Section 4.1.2) and these differences in focus and approach

contribute to the wide range of estimates, which for some S/R pairs disagree in sign and span an order of magnitude. In addition, prior efforts adopted varied regional definitions, metrics and periods of study, making it difficult to draw meaningful, quantitative estimates from a literature survey. By adopting a single approach across all models the HTAP intercomparison limits the factors contributing to differences in individual model estimates to treatment of emissions, chemistry, transport and resolution.

Estimates of intercontinental transport from the HTAP intercomparison for surface O3

The annual and spatial mean surface O3 decreases in each of the receptor regions resulting from 20% reductions of anthropogenic O3 precursor emissions for each source region are given in Table 4.2, along with the range of HTAP model predictions expressed in terms of the standard deviation between the different models. As expected the largest changes occur in the source regions.

For example, a 20% change in the anthropogenic emissions in North America (NA) leads to a mean change in surface O3 of ~1 ppbv, and these source region responses vary from a high of 1.26 ppbv in South Asia to 0.82 ppbv in Europe. The results also show that substantial changes in surface O3 occur far away from the source regions. For example, the annual mean surface O3 mixing ratio in Europe changes by 0.37 ppbv when current emissions in North America are changed by 20%.

To quantify the importance of changes in emissions outside the receptor area on surface O3

within the receptor region, we define the relative annual intercontinental response metric. This represents the ratio of the response in a particular region due to the combined influence of sources in the three other regions to the response from all source regions. This metric varies from 0% (receptor response entirely due to receptor region emissions) to 100% (receptor response entirely due to emissions elsewhere), with 50% representing the point at which the response from emissions outside the region is equal to the response from receptor region emissions. The relative annual

144

intercontinental response for annual average surface O3 over each region is shown in Table 4.2, and varies from 43% for Europe, to 32% for NA and SA. This indicates that in all four regions, emission changes in the three other source regions are 50-75% as important as emission changes over the receptor region itself.

Table 4.2. Annual and spatial mean surface O3 response (ppbv) to 20% decreases in anthropogenic precursor emissions (NOx, CO, NMVOC, plus aerosols and their precursors). Values are mean (median) ± one standard deviation across the 15 models that conducted the regional perturbation simulations (SR6). Bold font denotes responses to foreign emission perturbations that are at least 10%

of the response to domestic emission perturbations. Also shown is the relative annual intercontinental response for each receptor region defined as the ratio of the total response in mean surface O3 due to changes in the other three source regions compared to that due to changes in all regions.

Receptor Region

Source Region NA EU EA SA

Annual mean decrease

NA 1.04(1.03)±0.23 0.37(0.37)±0.10 0.22(0.24)±0.05 0.17(0.19)±0.04 EU 0.19(0.18)±0.06 0.82(0.68)±0.29 0.24(0.24)±0.08 0.24(0.25)±0.05 EA 0.22(0.23)±0.06 0.17(0.17)±0.05 0.91(0.86)±0.23 0.17(0.17)±0.05 SA 0.07(0.07)±0.03 0.07(0.07)±0.03 0.14(0.13)±0.03 1.26(1.18)±0.26 Relative annual intercontinental response

32% 43% 40% 32%

The spatial distribution of the changes in mean surface O3 levels due to 20% changes in anthropogenic precursors for each of the source regions in springtime are shown in Figure 4.7. As noted above, the largest changes in O3 occur over the source regions, but the influence is shown to extend throughout the northern hemisphere. These results indicate that a 20% decrease in North American anthropogenic emissions decreases mean surface O3 across the northern hemisphere by 0.1-0.5 ppbv. The decrease in European emissions has a marginally smaller impact, reducing mean surface O3 over North America by less than 0.35 ppbv and by 0.1-0.5 ppbv over Asia. The impact from a 20% decrease in East Asian anthropogenic emissions is slightly smaller, generally less than 0.3 ppbv except over western North America. The response to emission reductions over South Asia is more localized, with a decrease of less than 0.1 ppbv across much of the northern mid-latitudes, reflecting greater export into the tropical free troposphere. Over the Arctic, the largest impacts on surface O3 are from European emissions (~0.4 ppbv) and the smallest are from South Asian emissions (<0.1 ppbv). The differences between the contributing models (not shown here) are appreciable, with one standard deviation of 20-50% of the mean value over both source and receptor regions. This variation, a measure of uncertainty in the estimates, is largest at northern mid-latitudes for European emissions, for which the standard deviation reaches ~0.2 ppbv, roughly the same order as the multi-model mean decrease. On an annual basis, the standard deviation is generally less than half of the multi-model mean decrease in surface O3.

145

Figure 4.7. Model ensemble mean surface O3 decrease in springtime (in ppbv) for the combined 20% emission reduction over each source region showing the spatial variation over each region and over the Arctic. Each row shows the responses from a particular source region.

The range of estimates across the models participating in the HTAP intercomparison is generally smaller than that obtained from a survey of the literature, see Table 4.3. For a direct comparison with previous studies which have taken differing approaches to estimating

intercontinental contributions, the O3 responses in the HTAP studies derived from 20% emission changes are scaled by a factor of 5 to represent 100% emission changes. The uncertainties associated with this extrapolation are discussed later in Section 4.2.11. The largest intercontinental influence occurs for North American emissions on European surface O3, where the response to foreign emissions is between 1.0 and 2.55 ppbv. The largest S/R relationships are for NA→EU; EU→SA;

EU→EA; and NA→EA. These numbers are significant when compared with the changes in surface O3 due to changes in emissions from domestic sources. The largest values in Table 4.3 from prior studies (e.g., over NA in summer and over EU and EA for annual mean values) were estimated from simulations in which anthropogenic emissions were set to zero, or in which O3 production throughout the tropospheric column over a source region was considered to represent the effect of emissions from the source region. None of the models in the HTAP intercomparison suggest estimates near the upper end of this range.

146

Table 4.3. Annual and seasonal mean estimates for the contribution to surface O3 (in ppbv) over the receptor region from anthropogenic O3 precursor emissions in foreign source regions, arranged from largest (first row) to smallest (last row). The top entry in each cell is the full range from the 15 models that participated in the SR6 set of HTAP simulations. The response to the 20% emission perturbation was multiplied by five to estimate the full contribution; the limitations of this approach are discussed in the text. Also shown (italics) is the full range from the published literature included in Table 5.2 in HTAP [2007] and Figure 11 in Fiore et al. [2009], supplemented with estimates from more recent studies [Lin et al., 2010b; West et al., 2009b; Zhang et al., 2009]

Source/Receptor Annual DJF MAM JJA SON

NA

→

^{EU 1.00-2.55} _2.-15. ^0.70-2.35 _0.4 ^1.05-3.10 _0.2 ^1.05-2.65 _-0.3-5. ^1.00-2.45 _-0.9

EU

→

^{SA 0.70-1.60 0.55-2.00 0.75-1.95 0.75-1.80 0.65-1.45}

EU

→

^{EA 0.60-1.85} _0.-5.4 ^{0.35-2.95 0.75-2.80} _3. ^0.40-1.15 _3. ^0.65-1.70

NA

→

^{EA 0.60-1.55} _0.2-4.5 ^{0.65-2.05 0.50-1.60 0.30-1.35 0.60-1.70}

EA

→

^{NA 0.50-1.55} _1. ^{0.50-1.80 0.65-2.05} _4. ^0.30-1.15 _1.-3. ^0.45-1.20

NA

→

^{SA 0.50-1.15 0.40-1.55 0.60-1.45 0.35-0.85 0.45-1.00}

SA

→

^{EA 0.50-1.10 0.50-1.30 0.55-1.25 0.30-0.90 0.45-1.05}

EU

→

^{NA 0.40-1.55} _0.2-0.9 ^{0.30-1.90 0.55-1.95 0.30-2.05 0.45-1.15}

EA

→

^{SA 0.40-1.45 0.40-2.30 0.30-1.05 0.30-0.95 0.25-1.85}

EA

→

^{EU 0.35-1.30} _0.8-7. ^{0.35-1.45 0.40-1.70 0.30-1.05 0.35-1.05}

SA

→

^{EU 0.20-0.75 0.20-1.00 0.20-0.85 0.15-0.55 0.05-0.70}

SA

→

^{NA 0.20-0.65 0.20-0.90 0.25-0.85 0.10-0.45 0.15-0.60}

Seasonal responses

The annual mean surface O3 responses presented in Table 4.2 mask a large seasonal variability in the response to emission perturbations, as is evident from Table 4.3. Surface O3

responses to 20% emission changes in individual models are shown as a function of month in Figure 4.8. There is significant variability in the calculated responses in both source and receptor regions which varies by season and location. However, there is good agreement between models on the strong seasonal variations. For example, the response of European surface O3 to North American emissions, which averages 0.37 ppbv over the year, varies between 0.47 ppbv in spring, when the effects of intercontinental transport are generally largest in the Northern Hemisphere, and 0.29 ppbv in summer when O3 production from domestic emissions is greatest. The smallest responses are from South Asian emissions over North America, which average 0.07 ppbv over the year but range from 0.10 ppbv in February to 0.03 ppbv in August. At northern mid-latitudes the intercontinental influence is largest in boreal spring, with a secondary peak in fall in some locations, and is smallest during

summer when southerly flow is more prevalent over these regions. For the South Asian region there is less variation and the seasonality is reversed, reflecting the transition between dry and monsoon

147

seasons. The combined response of surface O3 over South Asia to foreign emissions varies from 0.4 to 0.9 ppbv, approximately 50% of the response due to domestic emissions reductions, and the

seasonality of domestic and intercontinental influences is similar, in contrast to the other regions considered here. Over the Arctic, the largest surface O3 response from all source regions occurs between April and June, with a secondary maximum in October and November, and this strong seasonality is most greatly influenced by European sources. These results indicate that while the intercontinental influence over many regions is important, decreasing domestic emissions is more effective at decreasing the highest O3 levels (e.g., as occur in July in EA, EU, and NA) [Fiore et al., 2009; Reidmiller et al., 2009].

Figure 4.8. Monthly mean surface O3 decreases (in ppbv) over receptor regions for the combined 20% emission reductions (HTAP SR6 simulations) showing (1) the seasonality of the responses (black line: ensemble mean ±1σ), and (2) the variability between models (grey lines). Each row shows the responses from a particular source region. For clarity, the vertical scale for the source regions (shaded panels, 0-2.5 ppbv) is different from that for the receptor regions (0-1 ppbv). The bottom row summarises the ensemble mean response over each

Figure 4.8. Monthly mean surface O3 decreases (in ppbv) over receptor regions for the combined 20% emission reductions (HTAP SR6 simulations) showing (1) the seasonality of the responses (black line: ensemble mean ±1σ), and (2) the variability between models (grey lines). Each row shows the responses from a particular source region. For clarity, the vertical scale for the source regions (shaded panels, 0-2.5 ppbv) is different from that for the receptor regions (0-1 ppbv). The bottom row summarises the ensemble mean response over each

Im Dokument Emission Inventories and Projections (Seite 165-0)